This study documents for the first time marked reductions in malaria cases and deaths in health facilities in two medium- and large-sized countries following large-scale distribution of LLIN and ACT. In particular in children under-five, declines of malaria cases and deaths were dramatic (~50% or higher), and occurred within one year of the scale-up of malaria interventions. This suggests that mass distribution of LLIN and nationwide roll-out of ACT conducted within months of the start of an expected malaria season may reduce cases and deaths during that transmission season.
In both Rwanda and Ethiopia, declines were of similar magnitude (~50% or higher) for in-patient cases and deaths, and out-patient laboratory-confirmed malaria cases. In Rwanda, all suspected malaria cases had laboratory examination. Since laboratory-confirmed diagnosis has better specificity for malaria than clinical diagnosis , this similarity supports the interpretation that apparent impacts are real, rather than being artifacts of changing patterns of diagnostic practices in the sampled facilities.
In Rwanda, the decline of in-patient and out-patient laboratory-confirmed malaria cases occurred in the face of increases in in-patient and out-patient non-malaria cases over 2001–2005, which probably reflected the introduction of health insurance schemes, resolution of civil conflict, and improvement of health services. The exceptionally low percentage decline in in-patient malaria deaths in Rwandans ≥ 5 years old was due to the doubling of reported malaria deaths in a single hospital (Nyanza) in the year 2007 (51 deaths, compared to < 25 annual malaria deaths during 2001–2006). As also non-malaria deaths in Nyanza hospital increased by 54% in 2007 compared to 2001–5, this sudden increase may relate to non-malaria factors. In any case, numbers of malaria deaths analysed (in both Rwanda and Ethiopia) were small, resulting in wide uncertainty ranges around mortality impact estimates.
In Ethiopia the strength of evidence was limited, first, by the small number of facilities with complete data for both age groups starting in 2001. Whereas the original sample had covered 13 out-patient and seven in-patient health facilities, only eight out-patient and five in-patient facilities had complete data over the period 2001–2007 (for at least all-ages combined) allowing inclusion in the present analysis. More fundamentally, the unstable nature of malaria transmission in Ethiopia  and the large year-to-year fluctuations in health facility burdens, also for non-malaria cases and deaths (Figure 4), make it impossible to draw firm conclusions yet regarding the causal relationship between the observed malaria declines and LLIN and ACT scale-up. Of note, Ethiopia experienced marked epidemics in 2003–4 [6, 7], after which declines in malaria indicators were expected even in the absence of intervention impact. Nevertheless, a repeat of the current analysis with a few additional years of post-intervention data could be expected to yield conclusive evidence as to the impact of the national LLIN and ACT scale-up.
The magnitudes of declines found in malaria indicators are in line with those reported from similar studies in small-scale areas of Kenya and Zanzibar. In three hospitals along the Kenya coast, in-patient pediatric malaria cases declined 28–63% as of March 2007, after distributions of ITNs and ACT . In District A in Zanzibar, in-patient malaria cases and deaths in children declined by 77% and 75%, respectively, within 24 months after the introduction of ACT in all 13 health facilities (and prior to substantial distribution of LLIN) .
The limitations were the following. Districts and health facilities were not randomly selected, but constituted a (stratified) convenience sample, selecting those sites where intervention scale-up had been relatively rapid and successful and where health facility data were of relatively good quality (e.g., in Ethiopia, excluding facilities where records did not go back to 2001). Therefore, estimated impacts cannot be extrapolated to the countries nation-wide. Also, while these results illustrate the benefits of rapid scale-up in the populations sampled, it would be inappropriate to extrapolate these findings to other countries with more intense malaria transmission, where interventions at similar coverage levels may have lower impact.
A more general limitation of health facility data is that they cover only the cases and deaths of patients who accessed the (public) health care system. Especially for ACT, it is possible that their coverage and impact is largely limited to the catchment populations of the facilities providing these drugs – with population-level impact diminishing by distance from health facilities. For this reason, it is difficult to extrapolate the observed health facility impacts to effects for the full populations living in the districts sampled. Inferences about population-level disease impact will typically require triangulation of several sources of data, including data from household surveys .
Third, in Ethiopia, indoor residual spraying (IRS) has been a well-established vector control intervention for a long period. It is applied in a focalized manner by targeting villages at risk for malaria epidemics. All districts visited had been applying IRS in a limited way during much of 2001–2007. The contribution of IRS to declines could not be estimated since IRS activities occurred throughout both pre- and post-intervention period.
Evaluation of impact is an essential part of modern programme management practice and will be needed by all high-burden African countries to meet the Roll Back Malaria goal of > 50% reduction in malaria-related mortality by 2010. Health facility data are important for quickly and continuously monitoring interventions with high impact at the health facility and district level. Facility-based surveillance may become even more important in the future, once wide-scale use of LLIN and ACT may change the epidemiology of malaria from stable endemic to unstable/epidemic. Areas of unstable and highly seasonal malaria will need to rely on continuous, timely surveillance to detect and respond to epidemics.
In conclusion, these initial data suggest that widespread distribution of LLIN and use of ACT in the public sector can result in marked reductions in the burden of non-severe and severe malaria morbidity and mortality seen in public-sector health facilities. International partners should urgently collaborate with national governments to ensure that all at-risk persons have access to appropriate vector control, including LLIN, and treatment with ACT. The magnitude of declines (≥ 50%) found in the studied facilities of Rwanda and Ethiopia was similar to that needed – at a population level – to reach the RBM target of 50% mortality reduction by 2010. It appears that marked reduction in malaria mortality can be achieved quickly and detected within a short time of scaling-up interventions, which may enable many African countries to make rapid progress towards the child survival targets in the Millennium Development Goals.